Theoretical and experimental analysis of the winner’s curse risk in RES-E auctions.

This report concludes the work carried out in the course of Task 3.4 of the AURES project. Its aim was to experimentally compare different auction design options that are relevant for renewable energy support (RES) under controlled conditions. The results complement theoretical findings and enable a better understanding of the effect of the design options.

The main research question in Task 3.4 was how different competition levels and price rules in multi-item procurement auctions influence the risk and the occurrence of the so-called winner’s curse, which refers to the phenomenon that a bidder suffers a loss after being awarded in the auction. This happens in a procurement auction when the award price is smaller than the actual costs to provide the good. This particularly applies to situations with high uncertainties for bidders regarding the costs of the auctioned good prior to the auction. Usually, those uncertainties correspond to so-called common value cost components, i.e., those that are similar or even the same for all participants in the auction.

The first auctions, where high common cost components were identified, were sales auctions for oil and gas leases (Capen, Clapp, & Campbell, 1971). In the context of RES, common cost components are especially prevalent in the case of offshore wind auctions, in which predefined and predeveloped projects are auctioned. Also in other RES auctions, common cost components play a key role as for example the PV module and wind turbine prices are similar for all bidders but unknown at the time of the auction due to the technological development.

The most common auction format for RES auctions are discriminatory or uniform price multi-unit procurement auctions. We extended the existing theory for auctions with common values and single-item supply (Milgrom & Weber, 1982; Kagel & Levin, 1986) for the respective multi-unit procurement auction equivalents. We introduce different competition levels through a variation in the number of auctioned goods. As a result, we derived the Bayes-Nash equilibria for both auction formats. In discriminatory price auctions, the equilibrium bidding strategy is the same for most cost estimations independent of the actual competition level. In the uniform price auction however, the equilibrium bidding strategies vary more. In general, the bidders bid more than their cost estimation to adapt for the winner’s curse but reduce this mark-up with a reduced competition level.

As a next step, we implemented an auction experiment under controlled laboratory conditions to test our theoretical predictions and to learn more about bidding behaviour in multi-unit procurement auctions with common costs. The three main results of this experiment can be summarized as follows: First, under both price rules, the subjects adjust their bids according to the different competition levels in a qualitatively correct but not quantitatively correct manner. That is, the bidders adapt their bids in a qualitatively right way (i.e. in the right direction) but they adapt too less to avoid the winner’s curse. Second and connected with the first, under both price rules, awarded bidders suffer from the winner’s curse with a high percentage. However, the occurrence and the magnitude of the winner’s curse decreases with a decreasing competition level, i.e., an increasing number of auctioned goods. Third, on an aggregate level we do not find significant differences between the two price rules with respect to the winner’s curse.

We conclude from our theoretical and experimental analysis that the winner’s curse risk matters in auctions for RES in case of high uncertainties of relevant common cost components. While this holds for discriminatory pricing and for uniform pricing, the winner’s curse risk increases when the competition level increases.

Download the report here.